Pipeline corrosion - assessing the damage

TWI Bulletin, November/December 1992

In 1985 he joined the Fatigue Department of TWI and worked on a wide range of problems, principally related to offshore structures. He is now Head of the Numerical Analysis section in the Engineering Department, responsible for finite element stress analysis activities.

Pipelines can suffer external or internal corrosion damage. Bob Andrews discusses methods of assessing the structural effects and presents some recent TWI work in this area.

Parent metal corrosion

For localised corrosion in carbon or HSLA steels there is an assessment procedure in the ASME piping code ^[1] usually referred to as the 'B31G' method. This is based on a fracture mechanics approach, and aims to restrict the membrane stress in the corroded area to the material flow stress. The flow stress is assumed to be 1.1 SMYS, where SMYS is the specification minimum yield strength. A formula is given which relates the allowable depth of corrosion to its axial length as a function of pipe wall thickness, diameter and operating pressure.

Recently proposals have been made to reduce the conservatism in the B31G approach. This proposal ^[2] known as 'RSTRENG', increases the assumed flow stress to SMYS + l0ksi and uses a more accurate and complex method of modelling the corroded profile. The increase in assumed flow stress is particularly beneficial for low strength material.

The B31G document specifically states that 'it should not be used to evaluate the remaining strength of corroded girth welds or longitudinal welds or related heat affected zones'. It also notes that it considers only effects of internal pressure and does not consider secondary stresses such as bending.

Although it is not explicitly stated, it is generally assumed that this method applies only to defects that are predominantly axial in orientation.

Some of these limitations are being addressed by finite element modelling of specific cases. For example, Chouchaoui et al ^[3] have carried out detailed elastic-plastic analyses of irregular corrosion defects, whilst Mok et al ^[4] have tested and modelled long spiral defects. Current work at Battelle ^[5] is aimed at producing an assessment method for combined pressure and bending loading. Work is also being carried out by British Gas ^[6] to investigate the interactions between adjacent areas of localised corrosion.

Assessment of corroded welds

Welds, both girth and longitudinal seam may be subject to preferential corrosion. Figure 1 shows some typical types of attack. Conventional fracture mechanics based assessment methods, such as PD6493:1991 ^[7] can be used directly if it is assumed that the groove is a sharp crack of the same depth. Both fracture and fatigue crack growth can be assessed on this basis. In many cases this approach may be essential for internal corrosion if it is not possible to determine with sufficient confidence whether an indication from NDE is crack-like or rounded. The conservative assumption is that it is a cracklike defect.

a) Root groove
b) Root HAZ corrosion
c) Corrosion at external cap toe

In some cases it may be known that corrosion of the weld is rounded; in these cases assuming a crack-like defect could be over conservative. Indeed, for smooth corrosion grooves it could be argued that attainment of adequate Charpy impact toughness will give a guarantee against brittle fracture.

The risk of fatigue failure from such a groove can be assessed by estimating the geometric stress concentration factor and then assessing the fatigue strength using an S-N approach, for example as in BS 5400: Part 10. ^[8] An S-N curve appropriate to a corroded surface and the external or internal environment should be used.

Where a corrosion groove intersects a pre-existing crack, or when a fatigue crack may initiate and grow from a groove, a stress intensity factor is required for fracture mechanics analysis. One method of obtaining this is by direct numerical analysis of the particular geometry. A simpler approach is to use established handbooks, such as Roark and Young ^[9] or Peterson ^[10] to estimate the stress concentration factor (SCF) for a groove in a strip, see Fig.2. This provides an estimate of the stress at the groove surface. For a Level 1 assessment to PD6493 this stress would be assumed to be the stress acting on the defect. At Level 2 it is permissible to take advantage of the decay in stress across the section, as sketched in Fig.2a. Glinka and Newport ^[11] have presented a method of estimating the decay of the stress field on moving away from a notch. This field can then be linearised over the defect length ( Fig.2b) and the resulting membrane and bending stresses used to estimate a stress intensity factor.

Localised corrosion

As noted above, the B31G ^[1] procedure is aimed at avoiding plastic collapse and hence considers only membrane stresses. By concentrating on membrane stresses this approach takes no account of elastic bending stresses which may be produced by a loss of section. Such additional stresses could act as a source of fatigue failure or initiate a fast fracture. Indeed, the additional bending stress together with the increase in membrane stress could cause a defect in a weld in or near the corroded area to become unsafe.

To assess these effects, preliminary finite element analyses were carried out at TWI for a typical offshore pipeline geometry with outside diameter 406mm and wall thickness 13mm. An operating pressure of 2000psi (13.8 N/mm ²) was used for the analyses, although the results can be scaled to any other pressure as the analyses were linear elastic.

a) Linearised across entire section
b) Linearised over the defect only

Two 2D models were created to simulate the extreme cases of a very long longitudinal loss of thickness, and a circumferential band around the entire pipe. In both cases losses of 10% and 20% of the wall thickness were modelled. For both depths a groove width of twice the thickness was modelled; for the 10% loss a groove width of four times the thickness was also analysed. Figure 3 shows typical finite element meshes; in both cases symmetry has been used to reduce the model to a half section with displacement constraints on the centreline.

a) Longitudinal groove 0.1t deep, 4t wide
b) Circumferential groove 0.2t deep, 2t wide

All analyses used version K4.1 of the MARC code. ^[12] For the very long groove ( Fig.3a) the analysis used eight noded isoparametric plane strain elements, whilst the analysis of the circumferential groove ( Fig.3b) used axisymmetric elements. The meshes were locally refined around the area of the groove; an initial check analysis showed that an 'ungrooved' model gave a uniform stress field. In both cases an internal pressure of 13.8 N/mm ² was applied. A further set of 3D analyses of the same pipe geometry was carried out to assess effects of constraint caused by a limited area of corrosion. Analyses used meshes as shown in Fig.4 using 20 noded isoparametric brick elements (MARC element 21). Circular patches of material loss of 10% and 20% were modelled, of diameters of two and four times the wall thickness. These were produced by programming a routine to move the nodes of an 'uncorroded' pipe to the required depth. This approach is now being extended at TWI to model irregular corrosion patches.

a) Mesh of pipe section
b) Detail of corroded patch 4t diameter, 0.2t deep

Figure 5 shows the results for longitudinal grooves and Fig.6 those for the circular patches, plotted as membrane and bending stress through the thickness at the centre of the corroded areas. Results for the circumferential groove are not shown, as the effects of pressure loading were small. The results show that under pressure loading, losses of thickness which would be accepted by B31G can produce significant elastic stress concentrations.

A 10% loss of wall is allowed regardless of axial length; the results in Fig.5 for the longitudinal groove show an increase in hoop membrane stress of 10%, as would be expected from the loss of section. An additional bending stress component of about 55 N/mm ² is produced. This may cause the combined stress to reach yield if the operating stress is at 72% of yield. More seriously, if fatigue or fracture assessments are carried out to assess a weld discontinuity, the additional stresses caused by corrosion could have a large effect.

When the thickness loss increases to 20%, the bending stress rises to 150 N/mm ². For this loss of thickness the axial length would be limited by B31G to 218mm, so the plane strain results are likely to be an over-estimate of the SCF. This is supported by the 3D model results, shown in Fig.6, where the hoop bending stress is only 49 N/mm ².

a) hoop stress
b) hxial stress

A bending stress of 150 N/mm 2 could cause a substantial loss in fatigue life if this acted on a longitudinal seam weld. It is unlikely that a properly operated pipeline would experience a significant number of full pressure fluctuations, so the practical significance of this result for fatigue loading must be limited. It is, however, possible that unstable fracture could be initiated from a pre-existing defect by these increased stresses.

Figure 6b also shows that pressure loading on the circular patch generates axial stresses, predominantly bending. This is believed to be due to bulging of the patch. Again, this additional stress may be significant if a defective girth weld had been assessed on the assumption of a low axial stress.

Detailed finite element analyses cannot be carried out for every defect. An attempt was made to estimate the elastic stresses using the formula given by Enquiry Case 79 of BS 5500:1991. ^[13] This gives a method of estimating the elastic SCF, K _t for misaligned pressure vessel welds. Full details are given elsewhere, ^[14] but the results showed both over and under estimates. Further work is required if a simple estimation method is to be developed.

Conclusions

Current methods of assessing the remaining strength of corroded pipelines have been reviewed and current research areas indicated. Methods of assessing locally corroded welds have been considered; there is a need for further research in this area. TWI will be launching a Group Sponsored Project on this topic. Readers are invited to contact the author for more details.

References